Spotting When AI Is Wrong and How to Fact-Check It

AI Basics · by Lung Tee · 5 reads

AI is incredibly capable, but sometimes it gives wrong answers with total confidence. This is called hallucination, or simply "making things up." Knowing how to spot it lets you use AI like a pro.

Why does AI get things wrong?

AI builds answers by predicting which words are likely to come next. It doesn't look up the truth every time. So when it lacks information, it may invent names, numbers, or events that sound convincing. Learn the basics in What Is Generative AI.

Where AI often goes wrong

Numbers, statistics, and complex calculations
Names of people, book titles, and references
Recent news and events
Law, taxes, and Thailand-specific information

Simple ways to check

Ask again from a different angle. If the answers don't match, dig deeper
Ask the AI which parts it's unsure about
Check against the official websites of the relevant organizations
Use tools that answer from your own documents, such as NotebookLM

Reduce errors from the start

Provide reference material along with your question
Instruct it: "If you don't know, say you don't know"
Break hard questions into steps. See Advanced Prompt Techniques

Fake images and videos

AI can create images and clips that look real. Before sharing shocking news or images, always check the source, and follow guidance from ETDA.

Be a smart AI user

Treat AI as a drafting assistant, not the final judge. Anything that affects money, health, or reputation should be double-checked every time. Read more about safe use in Using AI Safely.

An example of fact-checking

Suppose the AI says, "This law was passed in 2017." Follow up with, "How confident are you, and where should I verify this?" Then open the relevant agency's website to confirm. If the information doesn't match, tell the AI and ask it to correct its answer. A few minutes of checking can prevent big mistakes.

Match the level of checking to the stakes

Everyday tasks, like post ideas: a quick read-through is enough
External communication, like articles: verify every fact
Anything with legal or financial consequences: check with official sources or experts

Frequently asked questions

Do newer AI models still make mistakes?

Less often, but it still happens, so always check.

Can I ask another AI to check?

It helps to a degree, but official sources are still the best judge.

Summary

AI can be wrong, but if you know how to spot and verify errors, it becomes an incredibly powerful assistant. Remember: draft with AI, check with humans.